Unsupervised clustering for pattern recognition of heating energy demand in buildings connected to district-heating network

This paper presents a novel framework for the identification of different consumption patterns of heating loads of buildings. The approach to analyzing the consumption data is carried out by a combination of unsupervised clustering models. Density based clustering is used for outlier detection in the original dataset and K-means for pattern recognition. The proposed framework is then applied to a real building connected to the district heating in Tartu (Estonia). Three main day-types are identified for the building as an outcome of the clustering process, with different patterns throughout these days. More than 60% of the analyzed Cluster Validation Indexes studied in this paper show that classifying the daily demand profiles in three clusters is the optimal classification.